Data study · our own record

The blind quarter

Our price-change wire will not tell you that Ruby's 200-minute plan went from $375 to $720. It knows. It holds both figures, each dated, each read from Ruby's own page. It stays quiet because it cannot tell you which day that happened — and saying a date it does not know would make the record a claim about our scanning schedule rather than about the vendor.

Blind window

80 days

From

2026-06-20

To

2026-09-08

Datasets affected

267 / 271

Moves, vendors'

70

Differences, ours

152

What the numbers are, exactly

Between 2026-06-20 and 2026-09-08 — 80 days — the SaaS scanner did not run. It runs daily now. Comparing the last snapshot before that window with the first one after it, 222 priced figures differ across 267 of our 271 datasets.

152 of those differences are not the vendors'. They are ours: a September audit found the record holding an annual-billed rate in a month-to-month field for dozens of plans, and correcting that changes the number without any vendor touching a price. Every one is logged in the corrections log, and every one is excluded from the table below. Counting them as price moves is the single easiest way to inflate a figure like this, and it is the mistake this page exists to avoid.

That leaves 70 genuine moves — 55 up, 15 down. Each is stated as what it is: a figure observed on one date and a different figure observed on another. Not "raised on the 8th of September". We do not know that, and neither does anyone else who was not watching.

The moves

The differences that were ours

Separated rather than hidden. These 152 figures also differ across the two snapshots, and a naïve count would report them as vendor price rises. They are records we corrected — most of them plans whose monthly field held an annual-billed rate.

Limits

• These are observations, not events. A figure differing between two dates means the price changed at some point between them. It does not mean it changed once, or that it did not move and move back.
• A vendor added during the window is not here. Nor is a plan that did not exist in the earlier snapshot: arriving is not moving, and treating it as a move is how a price wire ends up announcing that a newly listed product "now lists" its first price.
• Some of the 70 may still be measurement, not price. The corrections log catches what we have found; a re-basing we have not yet found would sit in this table looking like a rise. Every row names its vendor and plan so it can be checked.
• This is one window, the widest. A dataset with several gaps contributes only its largest.

• blind-quarter.json — the full study, machine-readable, CC BY.
• Markdown twin — the same finding at a fraction of the tokens.
• Price changes — the dated wire, for moves we can date.
• Corrections — every figure we have changed, and why.